Switched on

A newly launched UK project will, it is hoped, spawn the next generation of super-fast adaptable communications networks. Christopher Sell reports.

A £20m DTI Technology Programme project was launched in March, bringing together four UK universities and some of the world’s leading multinationals.

Known as the Heterogeneous IP Networks (HIPNet) project, it will focus on the design and modelling of next-generation Information and Communication Technology (ICT) Networks.

ICT contributes a significant proportion to the UK’s GDP and is vital to the country’s economy, according to Dr Ken Guild of EssexUniversity, one of the principal researchers on the project. In the past five years the UK has developed one of the most competitive telecommunications markets in the world, with 3G services beginning to make an impact and plans being drawn up for the introduction of 4G and ultra wide-band wireless systems.

Guild said that ubiquitous real-time multi-media communications will dramatically increase the requirement for high-speed access and transport.

The scale of these next-generation, packet-based networks and the interactions between them will produce data behaviour that is difficult to model or predict.

The HIPNet project will develop the vision for a future telecommunications network landscape, taking into account the overall technological environment driving the evolution of converged IP-based networks and convergence of the application-communication sector over the next few years.

In so doing, it will ensure that the UK maintains its technological lead, by providing knowledge and skills in the design, modelling and verification of complex ICT networks through experimental research and modelling using a range of field and laboratory test-beds.

Importantly, it will look to model behaviour of networks and develop rules so that those built in the future can meet the evolving demands. This, according to Guild, is the key to the project. ‘It’s really all about bandwidth management. If you recall September 11 and the London bombings in July last year, both times the network crashed under the strain,’ he said.

‘The work we are doing will look to be able to send bandwidth where it is needed most.’

This essentially means devising a way to create a more flexible network that can be manipulated to cope with varying demands.

‘It is about understanding how a network will respond to unpredictable events such as terrorist attacks, traffic congestion or major news. As mobile devices become more sophisticated and the bandwidth to the end-user increases, the amount of “churn” or traffic variation will increase and this will put tremendous strain on both our current wired and wireless network infrastructures.’

The task is compounded by the varying performance requirements of the applications. Voice and video, said Guild, require a very stable constant bandwidth connection, whereas a web page or e-mail download can tolerate bandwidth variations from time to time. ‘It is the unpredictable nature of this traffic quantity, content and location that makes traffic engineering so difficult today.

‘To make matters worse,’ he said, ‘the heterogeneity is not only at traffic level, but also at the equipment and technology level and these are some of the issues that will be addressed within HIPNet.’ EssexUniversity is leading the future network demonstrator as well as the future network modelling work and will primarily focus on the transportation of traffic originating from e-science, e-health and grid applications. These include transferring X-ray images across the network to distributed radiologists and shifting high quantities of data in real time to enable remote surgery or scientific visualisation.

‘Transfer rates will vary depending on the type of application, but essentially our research will focus on rates in the range of 1-40Gbit/s,’ said Guild. ‘Part of our research [at Essex] will focus on very agile optical transport networks which are rapidly reconfigurable and use mainly optical processing/switching wherever possible.

‘A single wavelength (or colour) within an optical fibre can typically transport 10 or 40Gbit/s and this bandwidth can be shared among many users or it can be dedicated to a single application.

‘Take the example of a large file transfer between distributed supercomputers in a future e-science application. Today bandwidth is typically leased from a network provider on a long-term contract so the peak transfer rate is limited.

‘Consider a future scenario where the super-computers have 10Gbit/s optical interfaces and are directly connected to an agile optical transport network that utilises optical switches to optically route wavelengths around the network.

‘Normally, if there is no necessity for connectivity to other locations, no connection through the network is established (and no cost incurred). When a transfer is required, an automatic request is made to the network and the optical connection rapidly established and the file transferred.

‘In this case, it would take theoretically less than a second to transfer the entire contents of a DVD. Once complete, the connection would be released and the network would be made available to other users.

‘Obviously, there are many other implications such as how to stream a file at 10Gbit/s and how to ensure that today’s transport protocols (such as TCP) do not restrict such rapid transfers, but this is one such application scenario.’

A number of leading telecommunications companies are collaborating with the universities. Ericsson, Freescale Semiconductor and Artysan will look to provide design expertise at the component, system and network levels covering core IP networks, long haul and submarine transport networks, metro, access and enterprise networks.

‘Ericsson will focus on how to model, design and manage current and near-term network architectures. The Universities of Essex and Swansea will model futuristic network scenarios that utilise all-optical switching and processing techniques; this will also include the development of advanced traffic routing algorithms,’ said Guild. He added that to effectively model a multi-service network, understanding the service traffic characteristics is essential.

During the three and a half year project, traffic models will be developed that will convey the demands placed on the system by futuristic applications and services.

The models will be used within the network modelling activities and by Freescale and Artysan, who will develop high-density traffic sources to be used in the final validation phase of the network demonstrators. This will allow an emerging PICMG specification that is being developed for small form factor, field-replaceable telecom chassis to be evaluated within the project.

The research will culminate in two network demonstrators. One will be at Ericsson, Coventry and will include a range of commercially available network equipment from various vendors, to assess the network design rules derived within the project.

The other will utilise dark fibre — existing fibre-optic cables leased from network service providers — to connect the Essex research networks with a research network in Cambridge.

This will demonstrate the benefits of a ‘dynamic hybrid electro-optic transport network’ to meet the demands of future applications.